Commit
·
2ae3a4d
1
Parent(s):
9186715
Create helper.py
Browse files
helper.py
ADDED
@@ -0,0 +1,295 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import spacy
|
2 |
+
|
3 |
+
from geopy.geocoders import Nominatim
|
4 |
+
import geonamescache
|
5 |
+
import pycountry
|
6 |
+
|
7 |
+
from geotext import GeoText
|
8 |
+
|
9 |
+
import re
|
10 |
+
|
11 |
+
from transformers import BertTokenizer, BertModel
|
12 |
+
import torch
|
13 |
+
|
14 |
+
|
15 |
+
# initial loads
|
16 |
+
|
17 |
+
# load the spacy model
|
18 |
+
spacy.cli.download("en_core_web_lg")
|
19 |
+
nlp = spacy.load("en_core_web_lg")
|
20 |
+
|
21 |
+
# load the pre-trained BERT tokenizer and model
|
22 |
+
tokenizer = BertTokenizer.from_pretrained('bert-base-uncased')
|
23 |
+
model = BertModel.from_pretrained('bert-base-uncased')
|
24 |
+
|
25 |
+
# Load valid city names from geonamescache
|
26 |
+
gc = geonamescache.GeonamesCache()
|
27 |
+
city_names = set([city['name'] for city in gc.get_cities().values()])
|
28 |
+
|
29 |
+
|
30 |
+
def flatten(lst):
|
31 |
+
"""
|
32 |
+
Define a helper function to flatten the list recursively
|
33 |
+
"""
|
34 |
+
|
35 |
+
for item in lst:
|
36 |
+
if isinstance(item, list):
|
37 |
+
yield from flatten(item)
|
38 |
+
else:
|
39 |
+
yield item
|
40 |
+
|
41 |
+
|
42 |
+
def is_country(reference):
|
43 |
+
"""
|
44 |
+
Check if a given reference is a valid country name
|
45 |
+
"""
|
46 |
+
|
47 |
+
try:
|
48 |
+
# use the pycountry library to verify if an input is a country
|
49 |
+
country = pycountry.countries.search_fuzzy(reference)[0]
|
50 |
+
return True
|
51 |
+
except LookupError:
|
52 |
+
return False
|
53 |
+
|
54 |
+
|
55 |
+
def is_city(reference):
|
56 |
+
"""
|
57 |
+
Check if the given reference is a valid city name
|
58 |
+
"""
|
59 |
+
|
60 |
+
# Check if the reference is a valid city name
|
61 |
+
if reference in city_names:
|
62 |
+
return True
|
63 |
+
|
64 |
+
# Load the Nomatim (open street maps) api
|
65 |
+
geolocator = Nominatim(user_agent="certh_serco_validate_city_app")
|
66 |
+
location = geolocator.geocode(reference, language="en")
|
67 |
+
|
68 |
+
# If a reference is identified as a 'city', 'town', or 'village', then it is indeed a city
|
69 |
+
if location.raw['type'] in ['city', 'town', 'village']:
|
70 |
+
return True
|
71 |
+
|
72 |
+
# If a reference is identified as 'administrative' (e.g. administrative area),
|
73 |
+
# then we further examine if the retrieved info is a single token (meaning a country) or a series of tokens (meaning a city)
|
74 |
+
# that condition takes place to separate some cases where small cities were identified as administrative areas
|
75 |
+
elif location.raw['type'] == 'administrative':
|
76 |
+
if len(location.raw['display_name'].split(",")) > 1:
|
77 |
+
return True
|
78 |
+
|
79 |
+
return False
|
80 |
+
|
81 |
+
|
82 |
+
def validate_locations(locations):
|
83 |
+
"""
|
84 |
+
Validate that the identified references are indeed a Country and a City
|
85 |
+
"""
|
86 |
+
|
87 |
+
validated_loc = []
|
88 |
+
|
89 |
+
for location in locations:
|
90 |
+
if is_city(location):
|
91 |
+
validated_loc.append((location, 'city'))
|
92 |
+
elif is_country(location):
|
93 |
+
validated_loc.append((location, 'country'))
|
94 |
+
else:
|
95 |
+
# Check if the location is a multi-word name
|
96 |
+
words = location.split()
|
97 |
+
if len(words) > 1:
|
98 |
+
# Try to find the country or city name among the words
|
99 |
+
for i in range(len(words)):
|
100 |
+
name = ' '.join(words[i:])
|
101 |
+
if is_country(name):
|
102 |
+
validated_loc.append((name, 'country'))
|
103 |
+
break
|
104 |
+
elif is_city(name):
|
105 |
+
validated_loc.append((name, 'city'))
|
106 |
+
break
|
107 |
+
|
108 |
+
return validated_loc
|
109 |
+
|
110 |
+
|
111 |
+
def identify_loc_ner(sentence):
|
112 |
+
"""
|
113 |
+
Identify all the geopolitical and location entities with the spacy tool
|
114 |
+
"""
|
115 |
+
|
116 |
+
doc = nlp(sentence)
|
117 |
+
|
118 |
+
ner_locations = []
|
119 |
+
|
120 |
+
# GPE and LOC are the labels for location entities in spaCy
|
121 |
+
for ent in doc.ents:
|
122 |
+
if ent.label_ in ['GPE', 'LOC']:
|
123 |
+
if len(ent.text.split()) > 1:
|
124 |
+
ner_locations.append(ent.text)
|
125 |
+
else:
|
126 |
+
for token in ent:
|
127 |
+
if token.ent_type_ == 'GPE':
|
128 |
+
ner_locations.append(ent.text)
|
129 |
+
break
|
130 |
+
|
131 |
+
return ner_locations
|
132 |
+
|
133 |
+
|
134 |
+
def identify_loc_geoparselibs(sentence):
|
135 |
+
"""
|
136 |
+
Identify cities and countries with 3 different geoparsing libraries
|
137 |
+
"""
|
138 |
+
|
139 |
+
geoparse_locations = []
|
140 |
+
|
141 |
+
# Geoparsing library 1
|
142 |
+
|
143 |
+
# Load geonames cache to check if a city name is valid
|
144 |
+
gc = geonamescache.GeonamesCache()
|
145 |
+
|
146 |
+
# Get a list of many countries/cities
|
147 |
+
countries = gc.get_countries()
|
148 |
+
cities = gc.get_cities()
|
149 |
+
|
150 |
+
city_names = [city['name'] for city in cities.values()]
|
151 |
+
country_names = [country['name'] for country in countries.values()]
|
152 |
+
|
153 |
+
# if any word sequence in our sentence is one of those countries/cities identify it
|
154 |
+
words = sentence.split()
|
155 |
+
for i in range(len(words)):
|
156 |
+
for j in range(i+1, len(words)+1):
|
157 |
+
word_seq = ' '.join(words[i:j])
|
158 |
+
if word_seq in city_names or word_seq in country_names:
|
159 |
+
geoparse_locations.append(word_seq)
|
160 |
+
|
161 |
+
# Geoparsing library 2
|
162 |
+
|
163 |
+
# similarly with the pycountry library
|
164 |
+
for country in pycountry.countries:
|
165 |
+
if country.name in sentence:
|
166 |
+
geoparse_locations.append(country.name)
|
167 |
+
|
168 |
+
# Geoparsing library 3
|
169 |
+
|
170 |
+
# similarly with the geotext library
|
171 |
+
places = GeoText(sentence)
|
172 |
+
cities = list(places.cities)
|
173 |
+
countries = list(places.countries)
|
174 |
+
|
175 |
+
if cities:
|
176 |
+
geoparse_locations += cities
|
177 |
+
if countries:
|
178 |
+
geoparse_locations += countries
|
179 |
+
|
180 |
+
return (geoparse_locations, countries, cities)
|
181 |
+
|
182 |
+
|
183 |
+
def identify_loc_regex(sentence):
|
184 |
+
"""
|
185 |
+
Identify cities and countries with regular expression matching
|
186 |
+
"""
|
187 |
+
|
188 |
+
regex_locations = []
|
189 |
+
|
190 |
+
# Country references can be preceded by 'in', 'from' or 'of'
|
191 |
+
pattern = r"\b(in|from|of)\b\s([\w\s]+)"
|
192 |
+
additional_refs = re.findall(pattern, sentence)
|
193 |
+
|
194 |
+
for match in additional_refs:
|
195 |
+
regex_locations.append(match[1])
|
196 |
+
|
197 |
+
return regex_locations
|
198 |
+
|
199 |
+
|
200 |
+
def identify_loc_embeddings(sentence, countries, cities):
|
201 |
+
"""
|
202 |
+
Identify cities and countries with the BERT pre-trained embeddings matching
|
203 |
+
"""
|
204 |
+
|
205 |
+
embd_locations = []
|
206 |
+
|
207 |
+
# Define a list of country and city names (those are given by the geonamescache library before)
|
208 |
+
countries_cities = countries + cities
|
209 |
+
|
210 |
+
# Concatenate multi-word countries and cities into a single string
|
211 |
+
multiword_countries = [c.replace(' ', '_') for c in countries if ' ' in c]
|
212 |
+
multiword_cities = [c.replace(' ', '_') for c in cities if ' ' in c]
|
213 |
+
countries_cities += multiword_countries + multiword_cities
|
214 |
+
|
215 |
+
# Preprocess the input sentence
|
216 |
+
tokens = tokenizer.tokenize(sentence)
|
217 |
+
input_ids = torch.tensor([tokenizer.convert_tokens_to_ids(tokens)])
|
218 |
+
|
219 |
+
# Get the BERT embeddings for the input sentence
|
220 |
+
with torch.no_grad():
|
221 |
+
embeddings = model(input_ids)[0][0]
|
222 |
+
|
223 |
+
# Find the country and city names in the input sentence
|
224 |
+
for i in range(len(tokens)):
|
225 |
+
token = tokens[i]
|
226 |
+
if token in countries_cities:
|
227 |
+
embd_locations.append(token)
|
228 |
+
else:
|
229 |
+
word_vector = embeddings[i]
|
230 |
+
similarity_scores = torch.nn.functional.cosine_similarity(word_vector.unsqueeze(0), embeddings)
|
231 |
+
similar_tokens = [tokens[j] for j in similarity_scores.argsort(descending=True)[1:6]]
|
232 |
+
for word in similar_tokens:
|
233 |
+
if word in countries_cities and similarity_scores[tokens.index(word)] > 0.5:
|
234 |
+
embd_locations.append(word)
|
235 |
+
|
236 |
+
# Convert back multi-word country and city names to original form
|
237 |
+
embd_locations = [loc.replace('_', ' ') if '_' in loc else loc for loc in embd_locations]
|
238 |
+
|
239 |
+
return embd_locations
|
240 |
+
|
241 |
+
|
242 |
+
def identify_locations(sentence):
|
243 |
+
"""
|
244 |
+
Identify all the possible Country and City references in the given sentence, using different approaches in a hybrid manner
|
245 |
+
"""
|
246 |
+
|
247 |
+
locations = []
|
248 |
+
|
249 |
+
# add all the identified country/cities results in a list
|
250 |
+
|
251 |
+
try:
|
252 |
+
|
253 |
+
# ner
|
254 |
+
locations.append(identify_loc_ner(sentence))
|
255 |
+
|
256 |
+
# geoparse libs
|
257 |
+
geoparse_list, countries, cities = identify_loc_geoparselibs(sentence)
|
258 |
+
locations.append(geoparse_list)
|
259 |
+
|
260 |
+
# flatten the geoparse list
|
261 |
+
locations_flat_1 = list(flatten(locations))
|
262 |
+
|
263 |
+
# regex
|
264 |
+
locations_flat_1.append(identify_loc_regex(sentence))
|
265 |
+
|
266 |
+
# flatten the regex list
|
267 |
+
locations_flat_2 = list(flatten(locations))
|
268 |
+
|
269 |
+
# embeddings
|
270 |
+
locations_flat_2.append(identify_loc_embeddings(sentence, countries, cities))
|
271 |
+
|
272 |
+
# flatten the embeddings list
|
273 |
+
locations_flat_3 = list(flatten(locations))
|
274 |
+
|
275 |
+
# acquire the unique country/city names (because it is possible that many different approaches will capture the same countries/cities)
|
276 |
+
flat_loc_list = set(locations_flat_3)
|
277 |
+
|
278 |
+
# validate that indeed each one of the countries/cities are indeed countries/cities
|
279 |
+
validated_locations = validate_locations(flat_loc_list)
|
280 |
+
|
281 |
+
# create a proper dictionary with country/city tags and the relevant entries as a result
|
282 |
+
locations_dict = {}
|
283 |
+
|
284 |
+
for location, loc_type in validated_locations:
|
285 |
+
if loc_type not in locations_dict:
|
286 |
+
locations_dict[loc_type] = []
|
287 |
+
locations_dict[loc_type].append(location)
|
288 |
+
|
289 |
+
return locations_dict
|
290 |
+
|
291 |
+
except:
|
292 |
+
|
293 |
+
# handle the exception if any errors occur while identifying a country/city
|
294 |
+
print(f"An error occurred while checking if a city or country exists")
|
295 |
+
return ""
|